r Human Brain Mapping 33:27–39 (2012) r Levodopa and the Feedback Process on Set-Shifting in Parkinson’s Disease Wing Lok Au, 1 * Juan Zhou, 2,3 Paulito Palmes, 4 Yih-Yian Sitoh, 5 Louis CS Tan, 1 and Jagath C Rajapakse 2,3,6 1 Department of Neurology, Parkinson’s Disease and Movement Disorders Centre, National Neuroscience Institute, Singapore 2 School of Computer Engineering, Bioinformatics Research Centre, Nanyang Technological University, Singapore 3 Computational and Systems Biology Program, Singapore-MIT Alliance, Singapore 4 Department of Research, National Neuroscience Institute, Singapore 5 Department of Neuroradiology, National Neuroscience Institute, Singapore 6 Department of Biological Engineering, Massachusetts Institute of Technology, Cambridge, MA, USA r r Abstract: Objective: To study the interaction between levodopa and the feedback process on set-shifting in Parkinson’s disease (PD). Methods: Functional magnetic resonance imaging (fMRI) studies were per- formed on 13 PD subjects and 17 age-matched healthy controls while they performed a modified card- sorting task. Experimental time periods were defined based on the types of feedback provided. PD subjects underwent the fMRI experiment twice, once during ‘‘off’’ medication (PDoff) and again after levodopa replacement (PDon). Results: Compared with normal subjects, the cognitive processing times were prolonged in PDoff but not in PDon subjects during learning through positive outcomes. The ability to set-shift through negative outcomes was not affected in PD subjects, even when ‘‘off’’ medica- tion. Intergroup comparisons showed the lateral prefrontal cortex was deactivated in PDoff subjects during positive feedback learning, especially following internal feedback cues. The cortical activations were increased in the posterior brain regions in PDoff subjects following external feedback learning, especially when negative feedback cues were provided. Levodopa replacement did not completely restore the activation patterns in PD subjects to normal although activations in the corticostriatal loops were restored. Conclusion: PD subjects showed differential ability to set-shift, depending on the dopa- mine status as well as the types of feedback cues provided. PD subjects had difficulty performing set- shift tasks through positive outcomes when ‘‘off’’ medication, and showed improvement after levodopa replacement. The ability to set-shift through negative feedback was not affected in PD subjects even when ‘‘off’’ medication, possibly due to compensatory changes outside the nigrostriatal dopaminergic pathway. Hum Brain Mapp 33:27–39, 2012. V C 2011 Wiley Periodicals, Inc. Key words: executive functions; dopamine; functional magnetic resonance imaging r r Contract grant sponsors: Ministry of Education (MOE) and the Agency of Science, Technology and Research (A*Star) of Singapore. *Correspondence to: Wing Lok Au, National Neuroscience Insti- tute, 11 Jalan Tan Tock Seng, Singapore 308433. E-mail: [email protected]Received for publication 31 March 2010; Revised 4 September 2010; Accepted 20 September 2010 DOI: 10.1002/hbm.21187 Published online 24 March 2011 in Wiley Online Library (wileyonlinelibrary.com). V C 2011 Wiley Periodicals, Inc.
Transcript
r Human Brain Mapping 33:27–39 (2012) r
Levodopa and the Feedback Process onSet-Shifting in Parkinson’s Disease
Wing Lok Au,1* Juan Zhou,2,3 Paulito Palmes,4 Yih-Yian Sitoh,5
Louis CS Tan,1 and Jagath C Rajapakse2,3,6
1Department of Neurology, Parkinson’s Disease and Movement Disorders Centre,National Neuroscience Institute, Singapore
2School of Computer Engineering, Bioinformatics Research Centre,Nanyang Technological University, Singapore
3Computational and Systems Biology Program, Singapore-MIT Alliance, Singapore4Department of Research, National Neuroscience Institute, Singapore
5Department of Neuroradiology, National Neuroscience Institute, Singapore6Department of Biological Engineering, Massachusetts Institute of Technology, Cambridge, MA, USA
r r
Abstract: Objective: To study the interaction between levodopa and the feedback process on set-shiftingin Parkinson’s disease (PD). Methods: Functional magnetic resonance imaging (fMRI) studies were per-formed on 13 PD subjects and 17 age-matched healthy controls while they performed a modified card-sorting task. Experimental time periods were defined based on the types of feedback provided. PDsubjects underwent the fMRI experiment twice, once during ‘‘off’’ medication (PDoff) and again afterlevodopa replacement (PDon). Results: Compared with normal subjects, the cognitive processing timeswere prolonged in PDoff but not in PDon subjects during learning through positive outcomes. Theability to set-shift through negative outcomes was not affected in PD subjects, even when ‘‘off’’ medica-tion. Intergroup comparisons showed the lateral prefrontal cortex was deactivated in PDoff subjectsduring positive feedback learning, especially following internal feedback cues. The cortical activationswere increased in the posterior brain regions in PDoff subjects following external feedback learning,especially when negative feedback cues were provided. Levodopa replacement did not completelyrestore the activation patterns in PD subjects to normal although activations in the corticostriatal loopswere restored. Conclusion: PD subjects showed differential ability to set-shift, depending on the dopa-mine status as well as the types of feedback cues provided. PD subjects had difficulty performing set-shift tasks through positive outcomes when ‘‘off’’ medication, and showed improvement after levodopareplacement. The ability to set-shift through negative feedback was not affected in PD subjects evenwhen ‘‘off’’ medication, possibly due to compensatory changes outside the nigrostriatal dopaminergicpathway. Hum Brain Mapp 33:27–39, 2012. VC 2011 Wiley Periodicals, Inc.
Keywords: executive functions; dopamine; functional magnetic resonance imaging
r r
Contract grant sponsors: Ministry of Education (MOE) and theAgency of Science, Technology and Research (A*Star) ofSingapore.
*Correspondence to: Wing Lok Au, National Neuroscience Insti-tute, 11 Jalan Tan Tock Seng, Singapore 308433.E-mail: [email protected]
Received for publication 31 March 2010; Revised 4 September2010; Accepted 20 September 2010
DOI: 10.1002/hbm.21187Published online 24 March 2011 in Wiley Online Library(wileyonlinelibrary.com).
VC 2011 Wiley Periodicals, Inc.
INTRODUCTION
Parkinson’s disease (PD) is a chronic progressive neuro-degenerative disorder with both motor and cognitive man-ifestations. Executive dysfunction is one of the cognitiveimpairments seen in PD patients even in early stages ofthe disease [Hietanen and Teravainen, 1986; Lees andSmith, 1983; Muslimovic et al., 2005]. In particular, theability to set-shift is impaired in PD subjects [Brown andMarsden, 1988a; Cools et al., 1984; Flowers and Robertson,1985; Lees and Smith, 1983; Taylor et al., 1986]. However,the mechanisms mediating set-shift deficits in PD subjectshave remained controversial. Cortical activations may beincreased or decreased in PD subjects [Monchi et al., 2007]and levodopa replacement does not necessarily restore thecognitive networks in PD subjects to normal [Jubault et al.,2009]. Some authors have suggested that the increase incortical activations observed in PD subjects represents thecompensatory changes [Dagher et al., 2001; Samuel et al.,1997] whereas others have suggested that it implies directinvolvement of the mesocortical dopaminergic substratesin mediating cognitive deficits in PD subjects [Monchiet al., 2007]. There is also evidence to suggest that set-shiftdeficits in PD subjects may be mediated via non-dopami-nergic pathways [Kehagia et al., 2009; Lewis et al., 2005].
Using functional neuroimaging modalities, it has beenshown that different areas in the frontal, parietal, and tem-poral regions may be activated during performance of aset-shift task [Wager et al., 2004], without necessarilyinvolving the caudate nucleus [Monchi et al., 2006]. How-ever, it is not known whether these activations were attrib-uted to the set-shift process per se or due to a result ofclosely related executive functions such as working mem-ory and the feedback process. There are considerable over-laps in brain areas activated in set-shifting tasks andworking memory tasks [Wager et al., 2004], such as themedial prefrontal cortex (PFC), superior and inferior parie-tal, medial parietal and premotor cortices. The activationpatterns observed during set-shifting may also be influ-enced by the feedback process inherent to the set-shift task[Monchi et al., 2001]. The ventrolateral prefrontal cortex(VLPFC), caudate nucleus, and thalamus were activated innormal subjects receiving negative feedback; the dorsolat-eral prefrontal cortex (DLPFC) was activated when eitherpositive or negative feedback was received; and the puta-men showed increased activity while matching after nega-tive but not after positive feedback in the Wisconsin CardSorting Task [Monchi et al., 2001].
To our knowledge, the interaction between levodopaand the feedback process on set-shifting is not well under-stood. Nevertheless, to perform a set-shift, there must firstbe attention attracted by an internal or external feedbackmechanism to a specific perceptual dimension. This is usu-ally followed by an internal monitoring process to ascer-tain which aspect of the dimension is ‘‘rewarded’’ andwhich is ‘‘punished’’ before a response selection is made[Robbins, 2007]. The set-shift process is thus influenced by
multiple facets of executive functions: error and feedbackmonitoring, reward processing, and working memory.These cognitive functions are dependent on the properfunctioning of the striatum and its dopaminergic projec-tions [Holroyd and Coles, 2002; Lewis et al., 2004; Mcclureet al., 2004; Shohamy et al., 2004]. Furthermore, based onthe computational models of basal ganglia feedback mech-anisms, it has been suggested that PD subjects are better atlearning through negative feedback during ‘‘off’’ medica-tion and that levodopa reverses this bias, making PD sub-jects more responsive to positive outcomes [Frank et al.,2004]. With this background, we hypothesize that PD sub-jects will show differential ability to set-shift, dependingon the dopamine status as well as the types of feedbackcues provided. We hypothesize that the cortical activationsobserved in PD subjects in set-shifting may be attributedto the interactions between dopamine and the feedbackmechanisms inherent to the set-shift tasks. In particular,we postulate that PD subjects will have difficulty perform-ing set-shift tasks through positive outcomes when ‘‘off’’medication and that their performances will improve afterlevodopa replacement. We also postulate that the ability toset-shift through negative feedback will not be affected inPD subjects (both ‘‘off’’ and ‘‘on’’ medication), possiblydue to compensatory changes outside the nigrostriatal do-paminergic pathway (i.e., the mesocortical dopaminergicsubstrates or non-dopaminergic pathways).
METHODS
Subjects
We recruited 13 clinically definite PD subjects (7 men, 6women, mean age 61.9 � 7.4 years) according to the diag-nostic criteria of Calne et al. [1992], and 17 age-matchedhealthy subjects (9 men, 8 women, mean age 60.5 � 9.2years). All were right-handed ethnic Chinese. PD subjectshad mild to moderate disease severity (Hoehn and YahrStage 1 to 3), with mean disease duration of 4.9 � 3.5years. All PD subjects were on levodopa (mean dose 322.3� 102.0 mg/day), with three subjects on bromocriptine(mean dose 26.3 � 26.5 mg/day), four subjects on ropinir-ole (mean dose 2.8 � 1.6 mg/day), four subjects on benz-hexol (mean dose 2.5 � 1.0 mg/day), two subjects onamantadine (mean dose 150.0 � 70.7 mg/day), six subjectson selegiline (mean dose 10 � 0 mg/day), and one subjecton COMT inhibitor (400 mg/day). Subjects with atypicalparkinsonism, dementia, psychiatric illness, severe motorfluctuation, color blindness, on dopamine blocking agents,or with contraindications to functional magnetic resonanceimaging (fMRI) scanning were excluded from the study.Normal subjects on medications that might exert a dopa-minergic effect were also excluded from the study. All PDsubjects underwent the fMRI experiment twice in a day:one after overnight withdrawal of antiparkinson medica-tion for at least 12 h (PDoff), and the other during ‘‘on’’medication (PDon) at 40 min after being served the usual
r Au et al. r
r 28 r
morning dose of levodopa (mean morning dose 92.3 �18.8 mg). The study was approved by the InstitutionEthics and Review Board and all subjects gave writteninformed consent.
Quantitative Motor Assessments and
Cognitive Tasks
Subjects were briefed on the scanning procedures andexperimental conditions, and allowed up to 30 min topractice on the cognitive tasks outside the scanner untiltheir performance had reached a plateau. All subjectsachieved at least 60% accuracy rates on the tasks duringthe practice session. The Unified Parkinson’s Disease Rat-ing Scale (UPDRS) motor score and quantitative motorassessments were acquired prior to each fMRI experiment.Details of methods and analysis of timed motor testingusing the basic element of performance (BEP) module
(Human Performance Measurement Inc., Texas) weredescribed elsewhere [Au et al., 2008]. In brief, subjectswere measured on the index finger tapping (FT) speed,alternating hand (AH) movement speed, finger tappingspeed between two targets separated by a distance of 30cm (MS), and visuomotor reaction speed (RT). The scoreson both sides were added, and the values of FT, AH, andMS were summed to give an overall upper limb motorperformance index (UL Index). The higher the UL index,the better was the motor performance.
The cognitive task was performed using a modificationof the Montreal Card Sorting Task [Monchi et al., 2006]. Inthe Modified Card Sorting Task (MCST), external feedbackwas not provided so that subjects performed the MCSTtasks through implicit (or internal) learning without exter-nal feedback guidance (see Fig. 1). In the MCST with feed-back (MCST-F), external feedback was provided similar tothe Montreal Card Sorting Task. However, instead of a
Figure 1.
Modified Card Sorting Task (MCST). Subjects were asked to
match the stimulus card at the bottom of the computer screen
to one of the four index cards displayed in the top half of the
screen. The sorting principle was derived from a comparison of
attributes (color, number, or shape) between the stimulus and
index cards. Only the control condition blocks were shown in
this figure, where an exact match existed between the stimulus
and one of the index cards. In the continuous shift blocks (not
shown in this figure), the stimulus card contained only one at-
tribute shared with one of the four index cards. The matching
attributes of consecutive trials varied in a random order, such
that a shift was implicitly given by the task. In the MCST trials,
external feedback was not provided after a response was made.
Subjects had to perform the tasks through implicit or internal
learning without external feedback guidance. In MCST-F trials,
external feedback was provided after each response. A green
tick was displayed when the response was correct, and a red
cross was shown when the response was incorrect. All error
trials were removed from the fMRI analysis (e.g., trial number
3). Four experimental time periods were defined for the remain-
ing correct trials: (1) MCST-PF, correct trials just after receiving
a positive external feedback, (2) MCST-NF, correct trials just af-
ter receiving a negative external feedback, (3) MCST-CR, correct
trials just after a correct response was made in MCST trials,
and (4) MCST-IR, correct trials just after an incorrect response
was made in MCST trials.
r Set-Shifting in Parkinson’s Disease r
r 29 r
change in screen brightness which might be implicit to thesubjects, a red cross was displayed on screen after anincorrect response and a green tick was displayed whenthe response was correct. The feedback was thus explicitand remained on screen for 0.5 s. The external feedbackserved as external visual cues to guide subjects on the taskperformance.
In both MCST and MCST-F, four index cards were dis-played in a row in the top half of the computer screen.Starting from the left, the screen showed one red triangle,two green stars, three yellow crosses, and four blue circles.On each classification trial, a stimulus card was presentedin the middle of the screen below the index cards. Subjectswere asked to match stimulus cards to the four indexcards, using one of the attributes: number, color, or shape;the sorting principle was derived from a comparison ofattributes between the stimulus and index cards.
The original Montreal Card Sorting Task was designedwith four test conditions: control (C), continuous shift (S),retrieval with shift (RS), and retrieval without shift (R)[Monchi et al., 2006]. We had adapted the same paradigmin our card-sorting task, and in this study considered onlythe C and S conditions in order not to include the cogni-tive planning component from set-shifting [Monchi et al.,2006, 2007). In the C condition, there existed an exactmatch between the stimulus and one of the index cards.This condition served as the baseline for perceptual,motor, identity matching, and response selection compo-nents inherent to the card sorting task [Lie et al., 2006]. Inthe S condition, the stimulus card contained only one at-tribute (color, number, or shape) shared with one of thefour index cards. Therefore, only a single response waspossible on each trial. The matching attributes of consecu-tive trials varied in a random order such that the shift wasimplicitly given by the task [Monchi et al., 2006]. Therewere eight trials in each block of C and S condition. Thecondition blocks appeared randomly for four times. Stimu-lus would remain on screen until a response was received.The maximum response time allowed for each trial was7.5 s. The MCST-F block appeared in random either beforeor after the MCST block.
FMRI Scanning
Structural three-dimensional (3D) MR scans of the wholebrain were acquired using a 3 Tesla whole-body MRI scan-ner (Achieva 3.0, Philips Medical Systems, Best, The Neth-erlands). One hundred and eighty axial slices of T1-weighted 3D anatomical images (MPRAGE sequence)were acquired (TR ¼ 6.7 ms, TE ¼ 3.0 ms, FOV ¼ 230 �230 mm2, matrix ¼ 256 � 256, thickness ¼ 0.9 mm, voxelsize ¼ 0.90 � 0.90 � 0.90 mm3). Functional images werethen obtained with a T2-weighted gradient echo, echo pla-nar imaging (EPI sequence, 36 contiguous oblique axial 3mm slices, TR ¼ 2,000 ms, TE ¼ 30 ms, FOV ¼ 230 � 230mm2, acquisition matrix ¼ 128 � 128, voxel size ¼ 1.8 �
1.8 � 3 mm3 with blood oxygenation level dependent[BOLD] contrast). Visual stimuli were projected on ascreen and the experiment was controlled by EPrime soft-ware, mediated by the scanner-compatible IFIS Systemoutside the scanning room. Subjects were asked to respondto the visual stimuli by pressing one of the four keys on akeyboard provided to indicate answers for ‘‘one,’’ ‘‘two,’’‘‘three,’’ and ‘‘four,’’ corresponding to one of the fourindex cards. The left middle finger corresponded to ‘‘onered triangle,’’ the left index finger to ‘‘two green stars,’’ theright index finger to ‘‘three yellow crosses,’’ and the rightmiddle finger to ‘‘four blue circles.’’
Statistical Analysis
Statistical analyses of quantitative motor parameters andbehavioral data were performed using SPSS version 11.Student’s t-tests were performed at a significance level of0.05. FMRI data were analyzed using the Statistical Para-metric Mapping software (SPM2/SPM5) and the standardprocedures [Friston et al., 1995b]. All functional imageswere first corrected for head movement by using least-squares minimization [Friston et al., 1995a] and then cor-rected for slice timing. After coregistration to the subject’s3D T1-weighted anatomical MR image, functional imageswere spatially normalized into the SPM standard spacewith the anatomical image as a guide. Images were thenresampled at 2 mm, using Sinc interpolation, andsmoothed with a 3D Gaussian kernel with FWHM ¼ 8mm to decrease spatial noise. For an individual subject,the signal changes in BOLD contrast associated with theperformance of tasks were assessed on a voxel-by-voxelbasis, using the general linear model and the theory ofGaussian fields as implemented in SPM. The multivariateregression analysis used canonical haemodynamicresponse function with time and dispersion derivatives asbasis function, and corrected for temporal and spatialautocorrelations in the fMRI data. For every single condi-tion, all error trials were removed from the fMRI analysis.The start time and the length of each correct trial were ex-plicitly included in the design matrix. The S minus C con-trast was generated to look for activated regions specific tocontinuous set-shifting without cognitive planning [Mon-chi et al., 2007]. Four experimental time periods based onthe feedback mechanism in action were defined: matchingfollowing positive external feedback (MCST-PF), matchingfollowing negative external feedback (MCST-NF), match-ing following positive internal feedback (i.e., after a correctresponse was made in MCST [MCST-CR]), and matchingfollowing negative internal feedback (i.e., after an incorrectresponse was made in MCST [MCST-IR]). Group analyseswere done using random effect analysis (RFX) imple-mented in SPM5. The analysis used full-factorial designconsisting of three factors with following levels: threegroups (Normal vs. PDoff vs. PDon); two matching types(S vs. C); and four response types (CR vs. IR vs. PF vs.
r Au et al. r
r 30 r
NF), for a total of 24 treatment combinations. We eval-uated the effects of levodopa on set-shifting through inter-group comparisons within each of the task conditions(MCST-PF, MCST-NF, MCST-CR, MCST-IR). The effects offeedback were analyzed through comparisons of contrastsbetween MCST and MCST-F within each subject group.Model parameters were estimated using ReML (RestrictedMaximum Likelihood). Significant haemodynamic changesfor each contrast were assessed using the t-statistical para-metric maps. We reported activations below a threshold ofP < 0.005 (uncorrected) for multiple comparisons corre-sponding to t > 2.59 above a cluster size of greater than 30voxels. Activations that reached P < 0.05 (corrected) wereindicated in the tables by footnotes ‘‘a’’ and ‘‘b’’ for FalseDiscovery Rate (FDR) and Family-wise Error (FWE) correc-tions, respectively. Locations of significant activationswere identified by anatomical automatic labeling (AAL)with cluster approach [Tzourio-Mazoyer et al., 2002].Labels with the highest percentage per cluster were cho-sen, excluding those labeled ‘‘outside’’ by AAL.
RESULTS
Quantitative Motor Parameters
As expected, normal subjects had better motor perform-ance than PD subjects (see Fig. 2). Levodopa replacementimproved the UL index within the PD group (P < 0.002).The corresponding UPDRS motor score also improvedfrom 21 � 9 to 10 � 5 (P < 0.001). Despite the differencesin motor performance, there were no significant differen-ces in visuomotor reaction speed across subject groups(Control group: 9.7 � 1.5 per second, PDoff: 9.0 � 1.3 persecond, PDon: 9.0 � 1.6 per second).
Task Accuracy and Response Times
The set-shift accuracy rates were comparable across subjectgroups, regardless of levodopa replacement. The presence ofexternal feedback cues improved the accuracy rates within Sconditions in PD subjects (PDoff, P < 0.005; PDon, P < 0.01)but not in the normal group. The response times, however,were unaffected by the presence of the external feedback cues.Intergroup comparisons showed a trend towards longerMCST-PF and MCST-CR response times in PDoff comparedwith normal subjects (Fig. 3A). The corresponding responsetimes of PDon subjects were comparable to those of normalsubjects. There were no significant differences in MCST-NFandMCST-IR response times across subject groups (Fig. 3B).
FMRI Data
The S minus C contrasts were obtained for all subjectgroups to identify the activation areas during set-shifting(Tables I and II). Overall, PD and normal subjects showedactivations in one or more of the following areas duringset-shifting: frontal, parietal, and temporal cortices. Activa-tion areas were fewer in matched trials following positiveoutcomes (MCST-PF, MCST-CR) than in matched trials fol-lowing negative outcomes (MCST-NF, MCST-IR).
Activations during set-shifting following
positive outcomes (MCST-PF and MCST-CR)
There was a paucity of activation areas in normal andPDon subjects but not in PDoff subjects, during MCST-PF(Table I). In MCST-CR, only the right DLPFC and thalamuswere activated in normal subjects, with an absence of activa-tion areas in the PD group (with and without medication).
Activations during set-shifting following
negative outcomes (MCST-NF and MCST-IR)
In MCST-NF, the right DLPFC was activated in normaland PDon subjects but not in PDoff subjects (Table II). Onthe other hand, the caudate nucleus was weakly activated inPDoff (t ¼ 2.9, P ¼ 0.002) but not in PDon and normal sub-jects. Posteriorly, normal subjects activated the parieto-tem-poral areas, mainly on the left side. PDoff subjects activatedmainly the midline structures such as the cingulate cortexand precuneus. PDon subjects activated the right posteriorparietal cortex (PPC) and bilateral temporal lobes. In MCST-IR, normal subjects activated the left DLPFC, together withstrategic areas over the left parieto-occipital region and leftsuperior temporal pole. Activation areas in PDoff subjectswere limited to the right insula only. PDon subjects showeddiffuse activations over frontal and posterior brain regions.
Effects of external feedback vs. internalfeedback learning
Normal subjects activated the corticostriatal loop, suchas the left medial PFC, right caudate, left thalamus, and
Figure 2.
Age-adjusted upper limb motor performance index amongst PD
and healthy subjects. Normal, healthy subjects; PDoff, PD sub-
jects during ‘‘off’’ medication, PDon, PD subjects after levodopa
replacement. Error bar ¼ Standard Error of Mean (SEM).
r Set-Shifting in Parkinson’s Disease r
r 31 r
right supplementary motor area (SMA) during set-shiftingthrough internal feedback (MCST minus MCST-F) (TableIII), whereas the presence of an external feedback cue(MCST-F minus MCST) activated more of the mesocorticalsubstrates (DLPFC, middle cingulate cortex [MCC], and
insula on the right side) (Table IV). PDoff subjects had dif-fuse cortical activations during set-shifting through exter-nal feedback. Following levodopa replacement, the corticalactivations were focused mainly over the anterior brainregions during set-shifting through external feedback.
TABLE I. Significantly activated regions in continuous shift (S) minus control (C) contrasts in set-shifting task:
MCST-PF (with positive feedback) and MCST-CR (with correct responses)
Area
Normal PDoff
x, y, z t-stats Cluster size x, y, z t-stats Cluster size
The PDon group had no significant activations.L, left; R, right; DLPFC, dorsolateral prefrontal cortex; VLPFC, ventrolateral prefrontal cortex; PCN, precuneus; MTG, middle temporalgyrus. Locations identified by anatomical automatic labeling with cluster approach. Peak threshold levels at P < 0.005 (uncorrected).Number denotes t-statistics. Cluster size ¼ number of activated voxels.
Figure 3.
Response times to make a correct trial following (A) positive reinforcement (MCST-CR, MCST-PF),
at P < 0.05, P < 0.01, and P < 0.001 respectively. Error bar ¼ Standard Error of Mean (SEM).
r Au et al. r
r 32 r
Intergroup comparisons during internalfeedback learning
Inter-group comparisons during MCST-CR showed cort-ical deactivations in the hypodopaminergic state (Fig. 4A).Compared with normal subjects, the deactivations in theright DLPFC were greater in PDoff (t ¼ 4.16, P < 0.0001)than in PDon subjects (t ¼ 3.71, P < 0.0001). Inter-groupcomparisons during MCST-IR showed greater activity inthe PD group compared with normal subjects, especially
when levodopa replacement was given. The left PPC andthe right middle temporal gyrus were activated more inthe PD group than in normal controls. Greater activity wasalso noted in PDon than in normal subjects over diffusecortical areas, including the right caudate nucleus. Withinthe PD group, the corticostriatal loop activity was restoredafter levodopa replacement: bilateral DLPFC, right puta-men, and right SMA. The posterior cingulate cortex andstriate cortex on the right side were activated more inPDoff than PDon subjects.
TABLE II. Significantly activated regions in continuous shift (S) minus control (C) contrasts in set-shifting task:
MCST-NF (with negative feedback) and MCST-IR (with incorrect responses)
Area
Normal PDoff PDon
x, y, z t-stats Cluster size x, y, z t-stats Cluster size x, y, z t-stats Cluster size
L, left; R, right; DLPFC, dorsolateral prefrontal cortex; VLPFC, ventrolateral prefrontal cortex; lateral OBF, lateral orbitofrontal cortex;medial PFC, medial prefrontal cortex; PMC, primary motor cortex; SMA, supplementary motor area; ACC, anterior cingulate cortex;MCC, middle cingulate cortex; SSC, somatosensory cortex; AG, angular gyrus; PPC, posterior parietal cortex; PCN, precuneus; STP,superior temporal pole; STG, superior temporal gyrus; MTG, middle temporal gyrus; ITG, inferior temporal gyrus; HC, hippocampus;PHG, parahippocampal gyrus; OC, occipital cortex. Locations identified by anatomical automatic labeling with cluster approach. Peakthreshold levels at P < 0.005 (uncorrected). Number denotes t-statistics. Cluster size ¼ number of activated voxels.aActivations that reach P < 0.05 with FDR corrections.bActivations that reach P < 0.05 with FWE corrections.
r Set-Shifting in Parkinson’s Disease r
r 33 r
Intergroup comparisons during external
feedback learning
The right VLPFC was deactivated in the hypodopami-nergic state during MCST-PF (activations in normal >PDoff, t ¼ 3.16, P ¼ 0.001; activations in PDon > PDoff, t¼ 3.38, P < 0.0001). Comparing PDon to normal subjects,no significant differences in activity were noted in thePFC. Intergroup comparisons during MCST-NF showedincrease activity in DLPFC without caudate activations inPDon compared with normal subjects (Fig. 4B). On theother hand, PDoff subjects showed activity in the caudatenucleus without activating the DLPFC. Besides the DLPFC,the areas that were deactivated in the PD group comparedto normal subjects included the anterior cingulate cortex,MCC, supramarginal gyrus, and inferior temporal gyrus.Increased activations were noted in the temporo-parieto-occipital lobe in PDoff compared with normal subjects,and in the lateral orbitofrontal cortex, angular gyrus, andmidline structures of PDoff compared with PDon subjects.
DISCUSSION
By introducing a card-sorting paradigm with and with-out external feedback cues, we were able to show themodulation effects of both levodopa and the feedbackprocesses on set-shifting. Our results showed that PD sub-jects during ‘‘off’’ medication were less efficient in per-forming set-shift tasks compared with normal controlsduring both internal and external positive feedback learn-
ing. Levodopa replacement improved the cognitive proc-essing speeds in our PD subjects without significantimprovement in task accuracy. On the other hand, theability to set-shift through negative outcomes was notaffected in PD subjects, both during ‘‘off’’ and ‘‘on’’ medi-cation states. Taken together, these findings are in accord-ance with the observations by Frank et al. [2004] in thatPD subjects during ‘‘off’’ medication are better at learningthrough errors whereas PD subjects after levodopareplacement are more responsive to positive outcomes.Although the improvement in bradykinesia following levo-dopa replacement may improve the overall task responsetime, there were no significant differences in the visuomo-tor reaction speeds across subject groups. That is, thespeed at which a subject initiated a motor response follow-ing a visual stimulus was relatively unaffected by levo-dopa. Hence, the changes in task response times reflecteddifferences in cognitive processing speeds and notimprovements in motor speeds alone. Our work based onin vivo fMRI experiments confirm the findings of Franket al. [2004] based on computational modeling and cogni-tive procedural learning tasks.
Studies have shown that PD subjects perform cognitivetasks better through external feedback rather than throughinternal attentional control [Brown and Marsden, 1988a,b;Fimm et al., 1994; Horstink et al., 1990; Hsieh et al., 1995].Likewise, PD subjects in our study showed improvementin task accuracy following external feedback learning with-out significant effects on the task response times. In addi-tion, our findings showed that the ability to set-shiftdepended not only on the functioning DLPFC but also on
TABLE III. Significantly activated regions in MCST minus MCST-F contrasts in set-shifting task
Areas
Normal PDon
x, y, z t-stats Cluster size x, y, z t-stats Cluster size
The PDoff group had significant activations only in the right DLPFC (x ¼ 48, y ¼ 12, z ¼ 46, t-stats ¼ 3.69, cluster size ¼ 43).Abbreviations as in Table I and II. FUG, fusiform gyrus; AMG, amygdala; RO, rolandic operculum; EC, extrastriate cortex. Locationsidentified by anatomical automatic labeling with cluster approach. Peak threshold levels at P < 0.005 (uncorrected). Number denotes t-statistics. Cluster size ¼ number of activated voxels.aActivations that reach P < 0.05 with FDR corrections.bActivations that reach P < 0.05 with FWE corrections.
r Au et al. r
r 34 r
the interaction between the PFC and the other corticalareas. These interactions were modulated by the dopaminestatus of the subjects as well as the types of feedback cuesprovided. Overall, normal subjects in our study activateda set of cortical areas in the frontal, parietal, and temporalregions during set-shifting, similar to those reported inother studies [Lie et al., 2006; Monchi et al., 2007; Naga-hama et al., 1996; Rogers et al., 2000; Wager et al., 2004].In addition, our study showed that the corticostriatal loops(medial PFC, SMA, caudate nucleus) were activated dur-ing internal feedback learning whereas the mesocorticalsubstrates (DLPFC, cingulate cortex, insula) were activatedin the presence of external feedback cues. Intergroup com-parisons showed the lateral prefrontal cortex was deacti-vated in PDoff subjects during positive feedback learning,especially following internal feedback cues. The corticalactivations were increased in the posterior brain regions inPDoff subjects following external feedback learning, espe-cially when negative feedback cues were provided. Levo-dopa replacement did not completely restore theactivation patterns of PD subjects to normal although mostactivations in the corticostriatal loops were restored.
Our study, in particular, showed that both PD and nor-mal subjects had reduced cortical activations duringmatching following positive outcomes (MCST-CR, MCST-PF), and increased cortical activations during matching fol-lowing negative outcomes (MCST-IR, MCST-NF). Whilethe increase in cortical activations has been associatedwith poor task performance [Wager et al., 2005] and theexploratory phase of feedback learning [Sailer et al., 2007],it may also be a compensatory mechanism triggered byindividual subjects to cope with the set-shift demands[Dagher et al., 2001; Samuel et al., 1997]. In particular, cort-ical activations were increased in our PD subjects com-pared with normal controls during MCST-NF withoutcompromising the task performance. Studies on primateshave shown a phasic rise in dopamine levels when areward is presented, and a phasic fall in dopamine levelswhen an error is made [Schultz et al., 1997]. The phasicchanges in dopamine levels in response to the differentfeedback cues will lead to different cortical activations,depending on whether the mesocortical or the nigrostriatalpathways are affected [Cohen and Frank, 2009; Frank,2005; Guthrie et al., 2009]. Our findings suggest that the
TABLE IV. Significantly activated regions in MCST-F minus MCST contrasts in set-shifting task
Abbreviations as in Table I, II, and III. SMG, supramarginal gyrus; SC, striate cortex. Locations identified by anatomical automatic label-ing with cluster approach. Peak threshold levels at P < 0.005 (uncorrected). Number denotes t-statistics. Cluster size ¼ number of acti-vated voxels.aActivations that reach P < 0.05 with FDR corrections.bActivations that reach P < 0.05 with FWE corrections.
r Set-Shifting in Parkinson’s Disease r
r 35 r
Figure 4.
Intergroup comparisons of fMRI activation patterns in (A)
MCST-CR, and (B) MCST-NF. A: Deactivations were noted in
the right DLPFC in the hypodopaminergic state during MCST-
CR. Compared with normal subjects, the deactivations were
greater in PDoff (t ¼ 4.16, P < 0.0001) than in PDon subjects
(t ¼ 3.71, P < 0.0001). B: In MCST-NF, greater activations were
observed in the PDoff group compared to normal subjects in
the caudate nucleus and posterior brain regions, without activat-
ing the DLPFC. On the other hand, greater activations were
observed in the PDon group compared to normal subjects in the
DLPFC, without caudate activations. Both PDoff and PDon groups
showed deactivations in the cingulate cortex, supramarginal gyrus,
and inferior temporal gyrus, when compared with normal subjects.
The significance values are given as color-coded t-statistics.
increase in cortical activations observed in MCST-NF ismediated via the mesocortical pathways [Cools et al., 2002;Mattay et al., 2002; Monchi et al., 2007] through phasicdecrease in dopamine levels following an error response[Schultz et al., 1997]. Dopamine deficiency in the mesocort-ical pathway leads to cortical disinhibition with a loss offocusing effect of neural activity in the frontal lobe (andhence increased cortical activations) [Cools et al., 2002;Mattay et al., 2002; Monchi et al., 2007]. Set-shift deficitsobserved in PDoff subjects in our study during learningthrough positive outcomes were explained by deactiva-tions in the lateral PFC as compared with normal subjects.In particular, the right DLPFC was deactivated in MCST-CR and the right VLPFC was deactivated in MCST-PF.The reduced activations in the lateral PFC may beexplained by nigrostriatal dopamine deficiency with anincrease in thalamocortical inhibition [Albin et al., 1989;Alexander et al., 1986; Cools, 2006; Owen et al., 1998].
When a mental shift in task strategies is required, suchas matching following an incorrect response (MCST-IR),we noted activations in the left DLPFC together with stra-tegic areas over the left hemisphere (angular gyrus, supe-rior temporal pole, occipital lobe). These corticalinteractions were absent in PDoff subjects, likely due to ni-grostriatal dopamine deficiency with reduced DLPFC acti-vation [Albin et al., 1989; Nagano-Saito et al., 2008].However, we noted PDoff subjects were able to compen-sate by activating the right insula to cope with the set-shiftdemands [Soros et al., 2007; Taylor et al., 2009]. Intergroupcomparisons indicated that the PD group was able to acti-vate greater resources in the PPC than normal subjects.Both PPC and insula have been reported to mediate atten-tional set-shifting [Fox et al., 2003; Sylvester et al., 2003].The involvement of these areas may suggest possible non-dopaminergic pathways in mediating set-shifting [Kehagiaet al., 2009; Lewis et al., 2005]. On the other hand,although activations in the corticostriatal loops (DLPFC,SMA, putamen) were restored in PD subjects followinglevodopa replacement, the activation patterns in PD sub-jects did not return completely to normal. There were dif-fuse cortical activations in PDon subjects in the PFC andposterior brain regions compared to either PDoff or nor-mal subjects. With no significant differences in task per-formance across subject groups in MCST-IR in our study,the increase in cortical activations in PDon subjects maysuggest a less efficient compensatory mechanism in PDsubjects after levodopa replacement. Possible explanationsmay include functional disconnectivity in the corticalregions as a result of oversaturating the relatively intactmesocortical dopaminergic networks with levodopa [Roweet al., 2008], i.e., the inverted U-shape dose responsebetween dopamine and cognitive function [Cools, 2006;Tunbridge et al., 2006; Williams-Gray et al., 2008]. Othershave suggested that it is the phasic changes in dopaminelevels, which modulate the cortical activations, and thelevodopa replacement likely blunted the phasic dopami-nergic response [Frank, 2005; Guthrie et al., 2009]. It is also
possible that the cognitive processes involved in this do-main are not subject to dopaminergic depletion at all[Kehagia et al., 2009; Lewis et al., 2005]. In any case, theincrease in cortical activations observed in PDon comparedwith normal subjects in MCST-IR, despite having compa-rable set-shift performance in our study, suggests a less ef-ficient functional network in PDon subjects to cope withthe set-shift demands [Fimm et al., 1994].
There were several limitations to our study. Our samplepopulation could be small and restricted to PD subjects whowere stable responders to levodopa treatment. There couldbe differences in brain activation patterns and learning strat-egies in treatment naive subjects and motor fluctuators [Kuli-sevsky, 2000; Kulisevsky et al., 1996]. Future studies couldevaluate the effects of levodopa treatment in these groups ofpatients. In our study, PD subjects performed the fMRIexperiments twice within one day. There were concerns thatfatigue and ‘‘learning effects’’ may confound the fMRI data.However, we did not notice any significant difference in thevisuomotor reaction time across subject groups, which is asurrogate measure of attention and concentration. In orderto minimize these effects, we had all subjects trainedadequately on the tasks to ensure that the learning hadreached a plateau before scanning. Moreover, the task condi-tions were presented at random order for both fMRI experi-ments so that ‘learning effects’ were kept to a minimum.
The order in which PD subjects were scanned was notcounterbalanced in our study due to local logistic require-ments for PD subjects to have both ‘‘off’’ and ‘‘on’’ scansperformed in a day. While we acknowledge that the orderin which PD subjects were scanned (‘‘off’’ medication fol-lowed by ‘‘on’’ medication) may offer significant confoundin terms of the ‘‘order effect’’ [Konishi et al., 2008], thechanges in activation patterns observed in our study fromPDoff to PDon were different from those reported in nor-mal subjects with initial versus subsequent shifts on theWisconsin Card Sorting Task [Konishi et al., 2008]. More-over, the activation patterns of PDon minus PDoff subjectswere different across task conditions, with correspondingbehavioral results consistent with those reported by Franket al. [2004]. Therefore, even though underlying ‘‘ordereffect’’ cannot be completely excluded, our findings dosuggest possible interactions between levodopa and thefeedback process on set-shifting. Nonetheless, the ‘‘ordereffect’’ will need further evaluation in future studies.
The use of dopamine agonists in some of our PD sub-jects may confound the interpretation of the results in thisstudy, since dopamine agonists may have an effect on ex-ecutive functions [Costa et al., 2009]. Fortunately, theplasma elimination half-lives of the dopamine agonistsused by our PD subjects were relatively short (3–8 h forbromocriptine, 3–6 h for ropinirole) [Foley et al., 2004],and the 12-h overnight withdrawal of antiparkinson medi-cation is generally acceptable to define the clinically ‘‘off’’state in PD subjects [Langston et al., 1992].
In conclusion, we observed differences in set-shift taskperformance and brain activation patterns across subject
r Set-Shifting in Parkinson’s Disease r
r 37 r
groups, modulated by the dopamine status of the subjectsand also the types of feedback provided. The ability to set-shift in subjects depended not only on the DLPFC [Goeland Vartanian, 2005; Nagahama et al., 1996; Wager et al.,2005], but also on the interactions between the PFC andthe posterior brain regions [Cole and Schneider, 2007;Hampshire and Owen, 2006; Nagahama et al., 1999; Rog-ers et al., 2000; Wager et al., 2004]. Our findings showedthat PD subjects had impaired set-shifting through positiveoutcomes during ‘‘off’’ medication, which was mediatedvia reduced lateral PFC activations. We also observed thatthe ability to set-shift through negative feedback was notaffected in PD subjects (both ‘‘off’’ and ‘‘on’’ medication),possibly due to compensatory changes outside the nigro-striatal dopaminergic pathway (i.e., the mesocortical dopa-minergic substrates or non-dopaminergic pathways) assuggested by the fMRI findings in our study.
ACKNOWLEDGMENTS
The authors thank Ms. Irene Seah for her assistance inrecruiting subjects and performing the quantitative motorassessments, Ms. Keren-Happuch E. Fan Fen for perform-ing the behavioral tasks, and Dr. Oury Monchi for hisadvice on the experimental design. Dr. Paulito Palmes issupported by funding from the Singapore MillenniumFoundation Limited.
REFERENCES
Albin RL, Young AB, Penney JB (1989): The functional anatomy ofbasal ganglia disorders. Trends Neurosci 12:366–375.
Alexander GE, DeLong MR, Strick PL (1986): Parallel organizationof functionally segregated circuits linking basal ganglia andcortex. Annu Rev Neurosci 9:357–381.
Au WL, Seah ISH, Lau PN, Tan LCS (2008): Quantitative assess-ment of motor disability in Parkinson’s disease: Upper limbmotor performance index. Mov Disord 23:S347.
Brown RG, Marsden CD (1988a): An investigation of the phenom-enon of "set" in Parkinson’s disease. Mov Disord 3:152–161.
Brown RG, Marsden CD (1988b): Internal versus external cuesand the control of attention in Parkinson’s disease. Brain 111(Part 2):323–345.
Calne DB, Snow BJ, Lee C (1992): Criteria for diagnosing parkin-sons-disease. Ann Neurol 32:S125–S127.
Cohen MX, Frank MJ (2009): Neurocomputational models of basalganglia function in learning, memory and choice. Behav BrainRes 199:141–156.
Cole MW, Schneider W (2007): The cognitive control network:Integrated cortical regions with dissociable functions. Neuro-image 37:343–360.
Cools R (2006): Dopaminergic modulation of cognitive function-implications for L-DOPA treatment in Parkinson’s disease.Neurosci Biobehav Rev 30:1–23.
Cools AR, van den Bercken JH, Horstink MW, van SpaendonckKP, Berger HJ (1984): Cognitive and motor shifting aptitudedisorder in Parkinson’s disease. J Neurol Neurosurg Psychiatry47:443–453.
Cools R, Stefanova E, Barker RA, Robbins TW, Owen AM (2002):Dopaminergic modulation of high-level cognition in Parkin-son’s disease: The role of the prefrontal cortex revealed byPET. Brain 125:584–594.
Costa A, Peppe A, Dell’Agnello G, Caltagirone C, Carlesimo GA(2009): Dopamine and cognitive functioning in de novo sub-jects with Parkinson’s disease: Effects of pramipexole and per-golide on working memory. Neuropsychologia 47:1374–1381.
Dagher A, Owen AM, Boecker H, Brooks DJ (2001): The role ofthe striatum and hippocampus in planning: A PET activationstudy in Parkinson’s disease. Brain 124:1020–1032.
Fimm B, Bartl G, Zimmermann P, Wallesch CW (1994): Differentmechanisms underly shifting set on external and internal cuesin Parkinson’s disease. Brain Cogn 25:287–304.
Flowers KA, Robertson C (1985): The effect of Parkinson’s diseaseon the ability to maintain a mental set. J Neurol NeurosurgPsychiatry 48:517–529.
Foley P, Gerlach M, Double KL, Riederer P (2004): Dopamine re-ceptor agonists in the therapy of Parkinson’s disease. J NeuralTransm 111:1375–1446.
Fox MT, Barense MD, Baxter MG (2003): Perceptual attentionalset-shifting is impaired in rats with neurotoxic lesions of poste-rior parietal cortex. J Neurosci 23:676–681.
Frank MJ (2005): Dynamic dopamine modulation in the basal ganglia:A neurocomputational account of cognitive deficits in medicatedand nonmedicated Parkinsonism. J Cogn Neurosci 17:51–72.
Frank MJ, Seeberger LC, O’Reilly RC (2004): By carrot or by stick:Cognitive reinforcement learning in Parkinsonism. Science306:1940–1943.
Friston KJ, Holmes AP, Worsley KJ, Poline JB, Frith CD, Fracko-wiak RSJ (1995b): Statistical parametric maps in functionalimaging: A general linear approach. Hum Brain Mapp 2:189–210.
Goel V, Vartanian O (2005): Dissociating the roles of right ventrallateral and dorsal lateral prefrontal cortex in generation andmaintenance of hypotheses in set-shift problems. Cereb Cortex15:1170–1177.
Guthrie M, Myers CE, Gluck MA (2009): A neurocomputationalmodel of tonic and phasic dopamine in action selection: Acomparison with cognitive deficits in Parkinson’s disease.Behav Brain Res 200:48–59.
Hampshire A, Owen AM (2006): Fractionating attentional controlusing event-related fMRI. Cereb Cortex 16:1679–1689.
Hietanen M, Teravainen H (1986): Cognitive performance in earlyParkinson’s disease. Acta Neurol Scand 73:151–159.
Holroyd CB, Coles MGH (2002): The neural basis of human errorprocessing: Reinforcement learning, dopamine, and the error-related negativity. Psychol Rev 109:679–709.
Horstink MW, Berger HJ, van Spaendonck KP, van den BerckenJH, Cools AR (1990): Bimanual simultaneous motor perform-ance and impaired ability to shift attention in Parkinson’s dis-ease. J Neurol Neurosurg Psychiatry 53:685–690.
Hsieh S, Lee CY, Tai CT (1995): Set-shifting aptitude in Parkin-son’s disease: External versus internal cues. Psychol Rep77:339–349.
Jubault T, Monetta L, Strafella AP, Lafontaine AL, Monchi O(2009): L-dopa medication in Parkinson’s disease restores activ-ity in the motor cortico-striatal loop but does not modify thecognitive network. PLoS One 4:e6154.
r Au et al. r
r 38 r
Kehagia AA, Cools R, Barker RA, Robbins TW (2009): Switchingbetween abstract rules reflects disease severity but not dopami-nergic status in Parkinson’s disease. Neuropsychologia 47:1117–1127.
Konishi S, Morimoto H, Jimura K, Asari T, Chikazoe J, YamashitaK, Hirose S, Miyashita Y (2008): Differential superior prefrontalactivity on initial versus subsequent shifts in naive subjects.Neuroimage 41:575–580.
Kulisevsky J (2000): Role of dopamine in learning and memory—Implications for the treatment of cognitive dysfunction inpatients with Parkinson’s disease. Drugs Aging 16:365–379.
Kulisevsky J, Avila A, Barbanoj M, Antonijoan R, Berthier ML,Gironell A (1996): Acute effects of levodopa on neuropsycho-logical performance in stable and fluctuating Parkinson’s dis-ease patients at different levodopa plasma levels. Brain 119(Part 6):2121–2132.
Langston JW, Widner H, Goetz CG, Brooks D, Fahn S, Freeman T,Watts R (1992): Core assessment program for intracerebraltransplantations (CAPIT). Mov Disord 7:2–13.
Lees AJ, Smith E (1983): Cognitive deficits in the early stages ofParkinson’s disease. Brain 106 (Part 2):257–270.
Lewis SJ, Dove A, Robbins TW, Barker RA, Owen AM (2004):Striatal contributions to working memory: A functional mag-netic resonance imaging study in humans. Eur J Neurosci19:755–760.
Lewis SJ, Slabosz A, Robbins TW, Barker RA, Owen AM (2005):Dopaminergic basis for deficits in working memory but notattentional set-shifting in Parkinson’s disease. Neuropsycholo-gia 43:823–832.
Lie CH, Specht K, Marshall JC, Fink GR (2006): Using fMRI todecompose the neural processes underlying the WisconsinCard Sorting Test. Neuroimage 30:1038–1049.
Mattay VS, Tessitore A, Callicott JH, Bertolino A, Goldberg TE,Chase TN, Hyde TM, Weinberger DR (2002): Dopaminergicmodulation of cortical function in patients with Parkinson’sdisease. Ann Neurol 51:156–164.
Mcclure SM, York MK, Montague PR (2004): The neural substratesof reward processing in humans: The modern role of fMRI.Neuroscientist 10:260–268.
Monchi O, Petrides M, Petre V, Worsley K, Dagher A (2001): Wis-consin Card Sorting revisited: Distinct neural circuits participat-ing in different stages of the task identified by event-relatedfunctional magnetic resonance imaging. J Neurosci 21:7733–7741.
Monchi O, Petrides M, Strafella AP, Worsley KJ, Doyon J (2006):Functional role of the basal ganglia in the planning and execu-tion of actions. Ann Neurol 59:257–264.
Monchi O, Petrides M, Mejia-Constain B, Strafella AP (2007): Cort-ical activity in Parkinson’s disease during executive processingdepends on striatal involvement. Brain 130:233–44.
Muslimovic D, Post B, Speelman JD, Schmand B (2005): Cognitiveprofile of patients with newly diagnosed Parkinson disease.Neurology 65:1239–1245.
Nagahama Y, Fukuyama H, Yamauchi H, Matsuzaki S, Konishi J,Shibasaki H, Kimura J (1996): Cerebral activation during per-formance of a card sorting test. Brain 119 (Part 5):1667–1675.
Nagahama Y, Okada T, Katsumi Y, Hayashi T, Yamauchi H, Sawa-moto N, Toma K, Nakamura K, Hanakawa T, Konishi J,Fukuyama H, Shibasaki H (1999): Transient neural activity in themedial superior frontal gyrus and precuneus time locked withattention shift between object features. Neuroimage 10:193–199.
Nagano-Saito A, Leyton M, Monchi O, Goldberg YK, He Y,Dagher A (2008): Dopamine depletion impairs frontostriatal
functional connectivity during a set-shifting task. J Neurosci28:3697–3706.
Owen AM, Doyon J, Dagher A, Sadikot A, Evans AC (1998):Abnormal basal ganglia outflow in Parkinson’s disease identi-fied with PET. Implications for higher cortical functions. Brain121 (Part 5):949–965.
Robbins TW (2007): Shifting and stopping: fronto-striatal sub-strates, neurochemical modulation and clinical implications.Philos Trans R Soc B Biol Sci 362:917–932.
Rogers RD, Andrews TC, Grasby PM, Brooks DJ, Robbins TW(2000): Contrasting cortical and subcortical activations pro-duced by attentional-set shifting and reversal learning inhumans. J Cogn Neurosci 12:142–162.
Rowe JB, Hughes L, Ghosh BC, Eckstein D, Williams-Gray CH,Fallon S, Barker RA, Owen AM (2008): Parkinson’s diseaseand dopaminergic therapy—Differential effects on movement,reward and cognition. Brain 131:2094–2105.
Sailer U, Robinson S, Fischmeister FP, Moser E, Kryspin-Exner I,Bauer H (2007): Imaging the changing role of feedback duringlearning in decision-making. Neuroimage 37:1474–1486.
Samuel M, Ceballos-Baumann AO, Blin J, Uema T, Boecker H,Passingham RE, Brooks DJ (1997): Evidence for lateral premo-tor and parietal overactivity in Parkinson’s disease during se-quential and bimanual movements. A PET study. Brain 120(Part 6):963–976.
Schultz W, Dayan P, Montague PR (1997): A neural substrate ofprediction and reward. Science 275:1593–1599.
Shohamy D, Myers CE, Grossman S, Sage J, Gluck MA, PoldrackRA (2004): Cortico-striatal contributions to feedback-basedlearning: Converging data from neuroimaging and neuropsy-chology. Brain 127:851–859.
Soros P, Marmurek J, Tam F, Baker N, Staines WR, Graham SJ(2007): Functional MRI of working memory and selective atten-tion in vibrotactile frequency discrimination. BMC Neurosci8:48.
Taylor AE, Saint-Cyr JA, Lang AE (1986): Frontal lobe dysfunctionin Parkinson’s disease. The cortical focus of neostriatal out-flow. Brain 109 (Part 5):845–883.
Taylor KS, Seminowicz DA, Davis KD (2009): Two systems ofresting state connectivity between the insula and cingulate cor-tex. Hum Brain Mapp 30:2731–2745.
Tunbridge EM, Harrison PJ, Weinberger DR (2006): Catechol-o-methyltransferase, cognition, and psychosis: Val158Met andbeyond. Biol Psychiatry 60:141–151.
Tzourio-Mazoyer N, Landeau B, Papathanassiou D, Crivello F,Etard O, Delcroix N, Mazoyer B, Joliot M (2002): Automatedanatomical labeling of activations in SPM using a macroscopicanatomical parcellation of the MNI MRI single-subject brain.Neuroimage 15:273–289.
Wager TD, Jonides J, Reading S (2004): Neuroimaging studies ofshifting attention: A meta-analysis. Neuroimage 22:1679–1693.
Wager TD, Jonides J, Smith EE, Nichols TE (2005): Toward a tax-onomy of attention shifting: Individual differences in fMRIduring multiple shift types. Cogn Affect Behav Neurosci5:127–143.
Williams-Gray CH, Hampshire A, Barker RA, Owen AM (2008):Attentional control in Parkinson’s disease is dependent onCOMT val 158 met genotype. Brain 131:397–408.
